New Techniques for Auto-regionalisation of Reservoirs

History matching and appraisal of reservoirs usually include a partitioning of the reservoir for each physical property that is to be sensitised. Different engineers or geo-scientists often regionalise the reservoir differently. However, the choice of regionalisation can significantly affect critical results such as reserves estimates. Two auto-regionalisation methods were developed. The first method utilises small scale dimensional models to generate new property arrays compatible with those from the original geo-modelling. This approach also provides the option to regionalise using statistical fitting algorithms and the method of Principal Variables. The second method employs user-tailored distance measures to generate regions automatically. In trials the results were superior to the results obtained by conventional methods. Fewer runs were required to obtain a history match, with the second method performing the best.